Power transmission line icing detection and ice melting method and device

Abstract

The invention belongs to the technical field of power transmission lines, and particularly relates to a power transmission line icing detection and ice melting method and device, and the method comprises the steps: firstly, recognizing and positioning an icing region in a power transmission line on-site fusion image according to the data fused by a visible light image sensor and a thermal imaging sensor; generating an ice melting strategy according to the data of the icing identification and positioning unit; then receiving the space parameter, the thickness and the ice melting strategy of each section of ice coating area generated by the ice melting strategy design unit, generating a control instruction of an external ice melting device, and enabling the ice melting device to move to the target ice coating area and implement the corresponding ice melting strategy; carrying out centralized recording and management on the data of the whole process of the ice melting operation; based on the method, the icing detection precision, the response speed, the energy consumption control and the safety guarantee are obviously superior to those of the prior art, and efficient, safe and intelligent processing of icing of the power transmission line is achieved.

Description

Technical Field

[0001] This invention relates to the field of power transmission line technology, specifically to a method and apparatus for detecting and de-icing icing on power transmission lines. Background Technology

[0002] Transmission lines are prone to icing in winter or under extreme weather conditions. When a large amount of ice accumulates on the surface of the conductor, it will significantly increase the mechanical load on the line, leading to increased conductor sag, abnormal stress on the towers, or even collapse. In severe cases, it may cause large-scale power outages, resulting in huge economic losses and social impact.

[0003] Traditional transmission line icing detection mainly relies on manual inspections or single-sensor monitoring, which suffers from problems such as detection lag, insufficient accuracy, and poor real-time performance. Furthermore, manual inspections are extremely dangerous under severe weather conditions. Meanwhile, existing de-icing methods, such as overall current de-icing and DC de-icing, generally suffer from high energy consumption, low de-icing efficiency, and significant impact on line safety, making it difficult to meet the requirements of smart grids for refined and safe operation and maintenance. With the development of artificial intelligence, sensing technology, and intelligent equipment, combining visible light and thermal imaging sensors with dedicated chips for data processing and intelligent decision-making has become an important direction for solving the transmission line icing problem. However, some existing intelligent detection and de-icing solutions fail to form a complete closed loop of detection, strategy generation, execution, and feedback, resulting in weak inter-system coupling and difficulty in guaranteeing both de-icing efficiency and safety. Therefore, this invention proposes a method and device for detecting and de-icing transmission line icing. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A method for detecting and de-icing icing on power transmission lines, comprising the following specific steps: S1, collects transmission line data and establishes an initial feature image through multi-source sensors; S2 is designed as an IS intelligent computing chip to realize icing detection and ice melting decision-making.

[0005] As a preferred embodiment of the transmission line icing detection and de-icing method of the present invention, the specific steps of S1 are as follows: S11, the target transmission line area is detected by a visible light image sensor to obtain optical image data of the transmission line. D i Simultaneously, thermal imaging sensors are used to detect the same area, obtaining thermal imaging image data of the transmission line. D t ; S12, will D i and Dt The data is input to the data acquisition unit for preprocessing. S13, finally according to the set weight parameters α and β By fusing the two types of data, a fused image of the transmission line at the site is obtained. D f .

[0006] As a preferred embodiment of the transmission line icing detection and de-icing method of the present invention, the specific steps of S2 are as follows: S21, designed as an icing recognition and localization unit, is used to receive fused images. D f As input, D f After processing by the icing area perception model integrated within the unit, the field fusion of transmission lines is identified and located. D f There are areas covered with ice; S22, use the ice melting strategy design unit to generate the ice melting strategy; S23, the ice melting execution unit receives the spatial parameters, thickness and ice melting strategy of each ice-covered area generated by the ice melting strategy design unit, generates control instructions for the external ice melting device, and moves the ice melting device to the target ice-covered area and implements the corresponding ice melting strategy. S24 uses a recording and feedback unit to centrally record and manage data throughout the entire ice melting operation process.

[0007] As a preferred embodiment of the transmission line icing detection and de-icing method of the present invention, the specific steps of S21 are as follows: S211, Design an RFDEM (Regional Feature Depth Extraction Module) module for regional feature depth extraction, which realizes diversified extraction and fusion of input icing features through a multi-branch parallel processing mechanism; S212, Design an FMIM (Feature Multi-scale Interaction Module) module, which uses a multi-scale icing region feature interaction mechanism to achieve hierarchical extraction and fusion of input features from the icing region; S213, Design a MIFD (Multiple Information Fusion Detector) to achieve accurate detection of ice-covered area location information through a four-level branch collaborative prediction mechanism; S214. Based on the RFDEM module, FMIM module, and MIFD detector, an Ice-covered Area Perception Model (IAPM) is constructed. This model receives fused images of the transmission line site. D f As input, it automatically identifies and locates. D f The ice-covered areas that exist in the middle.

[0008] A transmission line icing detection and de-icing device, comprising: Visible light image sensors are used to detect target transmission line areas and obtain optical image data of the transmission lines; Thermal imaging sensors are used to detect target transmission line areas and obtain thermal imaging image data of the transmission lines. The IS intelligent computing chip is used to detect icing and make decisions on de-icing based on data from visible light image sensors and thermal imaging sensors.

[0009] In a preferred embodiment of the transmission line icing detection and de-icing device of the present invention, the IS intelligent computing chip includes: The icing identification and positioning unit is used to identify and locate areas with icing in the fused image of the power transmission line based on the data fused from the visible light image sensor and the thermal imaging sensor. The ice-melting strategy design unit is used to generate an ice-melting strategy based on the data from the ice-covering identification and positioning unit. The ice melting execution unit is used to receive the spatial parameters, thickness and ice melting strategy of each ice-covered area generated by the ice melting strategy design unit, generate control commands for the external ice melting device, and make the ice melting device move to the target ice-covered area and implement the corresponding ice melting strategy. The recording and feedback unit is used to centrally record and manage data from the entire ice melting operation process.

[0010] Compared with existing technologies: Compared to existing technologies, this invention proposes an IS (Ice-Sense) intelligent computing chip. This chip integrates an ice-covering recognition and positioning unit, an ice-melting strategy design unit, an ice-melting execution unit, and a recording and feedback unit, enabling intelligent detection, strategy generation, and efficient ice-melting operations for transmission lines. In the ice-covering recognition and positioning unit, this invention innovatively designs an Ice-Covering Area Perception Model (IAPM), effectively improving the recognition accuracy and robustness of ice-covered areas under complex lighting, low temperatures, and background interference environments, solving the problem of insufficient accuracy in traditional image detection methods. Simultaneously, the ice-melting strategy design unit in the IS chip can automatically generate targeted and differentiated ice-melting schemes by combining real-time acquired thickness, temperature, and regional distribution information, thereby reducing energy consumption and avoiding equipment risks caused by large-scale heating. The ice-melting execution unit drives a mobile execution device to precisely melt ice in the target area, improving operational efficiency. The recording and feedback unit realizes the storage and remote uploading of the entire operation data, constructing a closed-loop intelligent monitoring system. Overall, this invention is significantly superior to existing technologies in terms of icing detection accuracy, response speed, energy consumption control, and safety assurance, achieving efficient, safe, and intelligent handling of icing on transmission lines. Attached Figure Description

[0011] Figure 1 This is a flowchart of the ice detection and de-icing equipment of the present invention; Figure 2 This is a structural diagram of the RFDEM module of the present invention; Figure 3 This is a structural diagram of the FMIM of the present invention; Figure 4 This is a structural diagram of the MIFD of the present invention; Figure 5 This is a structural diagram of the IAPM model of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0013] This invention provides a method for detecting and de-icing icing on power transmission lines. Please refer to [link / reference]. Figures 1-5 The specific steps are as follows: S1 collects transmission line data and establishes an initial feature image through multi-source sensors.

[0014] The specific steps of S1 are as follows: S11, the target transmission line area is detected by a visible light image sensor to obtain optical image data of the transmission line. D iSimultaneously, thermal imaging sensors are used to detect the same area, obtaining thermal imaging image data of the transmission line. D t ; S12, will D i and D t The data is input to the data acquisition unit for preprocessing: First, a noise suppression algorithm is used to smooth the image and remove isolated noise points and environmental interference; then, the two types of images are registered in spatial coordinates so that the optical and thermal images maintain a correspondence at the position of the conductor. S13, finally according to the set weight parameters α and β By fusing the two types of data, a fused image of the transmission line at the site is obtained. D f : (1)

[0016] D f As input to subsequent icing identification and positioning units, it enables the comprehensive utilization of multi-source information. and β The optimal value needs to be approximated through repeated experiments.

[0017] S2 is designed as an IS intelligent computing chip to realize icing detection and ice melting decision-making.

[0018] Design an IS intelligent computing chip to fuse images of power transmission lines obtained from sensors. D f The input chip's internal icing detection and positioning unit identifies and calibrates the icing. D f The ice-covered area in the chip; then, the ice-melting strategy design unit within the chip is used to... D f The chip assesses the thickness and level of the icing area and generates a corresponding de-icing strategy. The de-icing execution unit in the chip then sends control commands to the external de-icing device, driving the heating device to carry out targeted de-icing operations in the target area. During the de-icing operation, the recording and feedback unit in the chip synchronously monitors temperature and image changes, records the de-icing operation in real time, and provides safety protection to ensure that the de-icing process is efficient, safe, and controllable.

[0019] The specific steps of S2 are as follows: S21, designed as an icing recognition and localization unit, is used to receive fused images. D f As input, D fAfter processing by the icing area perception model integrated within the unit, the field fusion of transmission lines is identified and located. D f There are areas covered with ice; The specific steps of S21 are as follows: S211, design an RFDEM (Regional Feature Depth Extraction Module) module for regional feature depth extraction. It achieves diversified extraction and fusion of input icing features through a multi-branch parallel processing mechanism. In the RFDEM module, the input feature R1 is first processed by the CBC module. Then, operations such as group convolution (GConv), deformable convolution (DConv), depthwise separable convolution (DSConv), deconvolution (TConv), and attention convolution (AConv) are performed on R1 respectively. The HADN+CELM operation is used for feature enhancement.

[0020] in: Group Convolution (GConv): Group convolution is a variant of convolution operation that reduces computation and the number of parameters by dividing the input channels into multiple groups and performing convolution operations on each group independently.

[0021] Deformable Convolution (DConv): Deformable convolution is a technique that extends traditional convolution. It can dynamically adjust the position of the convolution kernel according to the characteristics of the input data, increasing the spatial deformation capability of the convolution operation.

[0022] Depthwise separable convolution (DSConv) is a convolution method that breaks down the traditional convolution operation into two steps: first, each input channel is convolved independently (depthwise convolution), and then the channel information is merged through 1×1 convolution (pointwise convolution). Compared with standard convolution, depthwise separable convolution significantly reduces the amount of computation and the number of parameters, improving efficiency, and is particularly suitable for devices with limited resources.

[0023] Deconvolution (TConv): Deconvolution is commonly used in generative tasks to transform low-resolution feature maps into high-resolution images by learning the backpropagation of convolution kernels.

[0024] Attention Convolution (AConv): Attention Convolution introduces an attention mechanism that dynamically adjusts the weights of the convolution kernel based on the importance of different parts of the input features, thereby highlighting features in important regions.

[0025] The RFDEM module captures spatial, channel, and temporal multidimensional features through a parallel combination of heterogeneous convolution and pooling operations. Furthermore, its multi-branch feature fusion strategy significantly enhances the model's ability to analyze complex icing features. The module structure is as follows: Figure 2 As shown.

[0026] S2111: Input the ice-covered region feature map R1 into the CBM (Conv+BN+Mish) module for basic feature extraction to obtain the ice-covered region feature map R2.

[0027] Conv+BN+Mish is a publicly available combination of network layers. Conv (convolution operation) is responsible for extracting input features, BN (batch normalization) normalizes the features to speed up convergence and stabilize training, and Mish is an activation function used to introduce non-linearity and improve the model's expressive power.

[0028] S2112: Perform the GConv operation on the icing region feature map R2 to obtain the icing region feature map R3.

[0029] S2113: Perform weighted pooling (WPooling) on ​​the icing region feature map R2, followed by DConv operation, and then HADN+CELM operation to obtain the icing region feature map R4.

[0030] S2114: Perform average pooling (APooling) on ​​the ice-covered region feature map R2, followed by DSConv operation, and then HADN+CELM operation to obtain the ice-covered region feature map R5.

[0031] S2115: Perform the TConv operation on the ice-covered region feature map R2, then process it through the HADN+CELM operation, and then perform average pooling (APooling) to obtain the ice-covered region feature map R6.

[0032] S2116: Perform AConv on the icing region feature map R2, and then process it using the HADN method to obtain the icing region feature map R7.

[0033] S2117: Concatenate R3, R4, R5, R6, and R7 according to their channels to obtain the final output feature map of the icing area, R8.

[0034] The HADN+CELM operation is a composite operation designed in this invention. In this operation, the input data is first normalized using HADN to adjust the data distribution and adapt it for subsequent operations. Then, the normalized data is nonlinearly mapped using the CELM function to further optimize the data representation. This operation combines the advantages of normalization and nonlinear mapping, enhancing the model's data processing capabilities and learning performance.

[0035] HADN (Hyper-Adaptive Dynamic Normalization): HADN is a normalization method that dynamically and adaptively adjusts the normalization parameters based on the features of the input data, thereby improving the model's adaptability to changes in the input data. In this way, HADN can better handle the unevenness of data distribution, thus improving the model's learning effectiveness and performance.

[0036] CELM (Cubic Exponential Logistic Mapping): CELM is a nonlinear transformation function that transforms input values ​​according to a specific exponential logistic curve through cubic exponential logistic mapping. This method is often used for data mapping and compression, effectively processing and transforming input data to adapt to different model requirements and complex tasks.

[0037] This invention designs a HADN (Hyper-Adaptive Dynamic Normalization) method to perform high-dimensional adaptive normalization and nonlinear dynamic adjustment on the input icy region features in the module, so as to improve the module's ability to express complex icy region feature distributions and training stability.

[0038] In the HADN normalization method, the input feature map is first... Calculate local statistics by channel, including the dynamically weighted mean. With dynamic weighted variance : (2) (3)

[0040] in, The learnable dynamic weights are derived from the subnetworks. The generation process can adaptively adjust the contribution of each pixel to the statistics based on local features. It is a very small positive number to avoid division by zero or gradient explosion / vanishing caused by the denominator being 0 or close to 0.

[0041] Then the input features are standardized to obtain And introduce a multi-channel nonlinear coupling mapping: (4)

[0043] in, For learnable nonlinear intensity parameters, This is the inter-channel coupling coefficient, which can adaptively adjust the influence of different channels on the current channel during training, thereby achieving cross-channel information sharing and dynamic enhancement.

[0044] The HADN method can perform adaptive weighting and nonlinear mapping on different channels and spatial locations while ensuring normalization characteristics, thereby enhancing the expressive power of important features, suppressing outliers, and dynamically optimizing parameters during training. This significantly improves the model's performance and stability in the task of processing images of icing on power transmission lines.

[0045] This invention designs a CELM (Cubic Exponential Logistic Mapping) function to perform complex nonlinear mapping on the feature map of icing areas, thereby improving the module's ability to perceive features of different icing areas and its training stability.

[0046] The CELM activation function is defined as follows: (5)

[0048] in, Input to neurons; Adjusting the magnitude of the cubic exponent term enhances the nonlinear response to positive inputs; Control the curve shape and output compression of the Logistic mapping part to avoid gradient explosion; control The suppression term achieves smooth decay of negative inputs and prevents dead neurons. The function provides nonlinear enhancement to positive and negative inputs through a cubic exponential term, smooths extremum outputs through a logistic mapping, and... The suppression term ensures the gradient continuity for negative inputs. The parameters of the activation function... It enables adaptive learning during training, allowing the model to automatically adjust the response strength and nonlinearity of positive and negative inputs according to task requirements. CELM achieves adaptive adjustment of input signals through multi-level nonlinear operations, significantly enhancing the model's expressive power and improving training stability.

[0049] S212 designs a Feature Multi-scale Interaction Module (FMIM). This module utilizes a multi-scale feature interaction mechanism for iced regions to achieve hierarchical extraction and fusion of input features from these regions. The FMIM module combines multi-scale parallel processing with serial refinement to achieve hierarchical modeling of the feature space of iced regions. Furthermore, it employs a lightweight combination of depthwise separable convolutions and pointwise convolutions to improve computational efficiency while maintaining performance. The module structure is as follows: Figure 3 As shown.

[0050] S2121: Perform a Conv operation on the icing region feature map F1 to obtain the icing region feature map F2.

[0051] S2122: Perform FSUpsampling upsampling on F2 to obtain the feature map F3 of the icy area.

[0052] S2123: Segment F1, F2, and F3 according to channels to obtain the icing area feature map F4.

[0053] S2124: Perform the DSConv operation on F4 to obtain the icing region feature map F5.

[0054] S2125: Perform DSConv operation on F4, followed by PConv (pointwise convolution) operation to obtain the ice-covered region feature map F6.

[0055] S2126: Perform the DConv operation on F4 to obtain the icing region feature map F7.

[0056] S2127: Perform the DSConv operation on F4, followed by the PConv operation, to obtain the icing region feature map F8.

[0057] S2128: Perform DConv on F4 to obtain the feature map F9 of the icing area.

[0058] S2129: Concatenate F5, F6, F7, F8, and F9 by channel, and then perform PConv to obtain the final output feature map of the icing area, F10.

[0059] This invention designs a Fast Spatial Upsampling (FSUpsampling) method for efficiently and accurately recovering the spatial resolution of feature maps of iced regions. The FSUpsampling method can be expressed by the following formula, ultimately obtaining each pixel in the feature map of the iced region. Generate upsampled pixel values : (6)

[0061] Among them, convolution kernel Used to extract local features of each pixel It is the length of the convolution kernel. It is the width of the convolution kernel. and The value is automatically learned through training a neural network model. During the training process, It adaptively optimizes based on image content to capture local texture, edges, and information at different scales, thereby improving upsampling performance; adaptive weights Used to adjust the contribution of each input pixel to the output pixel. It is generated by a convolutional subnetwork that dynamically adjusts the weight of each pixel based on the local context information of the input image, ensuring that more attention is given to edge and textured regions, while contributing less to flat regions.

[0062] The advantage of the FSUpsampling method is that it allows the upsampling of each pixel to depend not only on surrounding local features but also on the overall image structure, ensuring that the upsampled image both restores details and avoids over-smoothing. This is achieved through convolutional kernels. With weight Through the combined effect of [various technologies], FSUpsampling achieves fine upsampling without relying on traditional interpolation or deconvolution operations, effectively restoring the details and textures of high-resolution images.

[0063] S213, a MIFD (Multiple Information Fusion Detector) is designed to achieve accurate detection of location information in icy areas through a four-branch collaborative prediction mechanism. The four-branch prediction architecture of the MIFD detector, through a heterogeneous combination of DSConv and standard Conv, enables complementary location features under different receptive fields. The detector structure is as follows: Figure 4 As shown.

[0064] S2131: Input the ice-covered area feature map M1 into the CBM (Conv+BN+Mish) module for basic feature extraction to obtain the ice-covered area feature map M2.

[0065] S2132: Perform Conv+HADN+CELM twice in parallel on M2, and then add the results of the two processes together to obtain the icing region feature map M3.

[0066] S2133: Perform the DSConv operation on M3 to obtain the icing region feature map M4.

[0067] S2134: Perform a Conv operation on M3 to obtain the icing region feature map M5.

[0068] S2135: Perform the DSConv operation on M3 to obtain the icing region feature map M6.

[0069] S2136: Perform a Conv operation on M3 to obtain the icing region feature map M7.

[0070] S2137: Perform WAdd (adaptive weight fusion) on M4, M5, M6, and M7 to obtain the final output location of the icing area, M8.

[0071] S214. Based on the RFDEM module, FMIM module, and MIFD detector, an Ice-covered Area Perception Model (IAPM) is constructed. This model receives fused images of the transmission line site. D f As input, it automatically identifies and locates. D f The icing-covered areas exist within. The structure of the IAPM model is as follows: Figure 5 As shown.

[0072] S2141: Fusion of images of transmission line sites D f Inputting the CBS layer yields the feature map I1 of the icing area of ​​the first transmission line.

[0073] S2142: Input the feature map I1 of the icing area of ​​the first transmission line into the RFDEM layer to obtain the feature map I2 of the icing area of ​​the second transmission line.

[0074] S2143: Input the feature map I2 of the icing area of ​​the second transmission line into the FMIM layer, and then perform DS (DownSampling) to obtain the feature map I3 of the icing area of ​​the third transmission line.

[0075] S2144: Input the feature map I3 of the icing area of ​​the third transmission line into the RFDEM layer to obtain the feature map I4 of the icing area of ​​the fourth transmission line.

[0076] S2145: Input the icing area feature map I4 of the fourth transmission line into the FMIM layer, and then perform DS to obtain the icing area feature map I5 of the fifth transmission line.

[0077] S2146: Input the feature map I5 of the icing area of ​​the fifth transmission line into the MIFD layer to obtain the final output of the icing area location.

[0078] S22, using the ice-melting strategy design unit, generates the ice-melting strategy; this unit is based on the calibrated ice-covering recognition and positioning unit. D f The spatial location, size, and thickness of each icing area in the image are calculated by combining the conductor geometry model, thermal imaging data, and environmental parameters (such as temperature, wind speed, and current load), and a targeted de-icing strategy is generated.

[0079] In the strategy design unit, the image coordinates of the icing area are first mapped to a real spatial coordinate system with specific latitude and longitude, obtaining the starting position, ending position, and length of each icing area; this is then combined with the thermal imaging temperature distribution. Wire diameter and environmental correction factor Through the thickness function Estimate the thickness of ice cover in each section. , The actual area of ​​the icing region is calculated by combining the pixel coverage area (calculated by mapping the number of pixels to the corresponding length and width in actual space). , For length, For width.

[0080] Then, a de-icing strategy is generated based on the icing thickness level, conductor material, and safety constraints. The de-icing strategy includes the following: The icing thickness grades of this invention are as follows: Thin ice areas: The ice thickness is relatively small, generally less than 5 mm.

[0081] Medium to thick areas: The ice thickness is relatively large, usually 5 mm or more (which can be further subdivided into moderate 5–10 mm, heavy 10–20 mm, and extra-heavy >20 mm).

[0082] 1. Selection of de-icing method: Select the appropriate de-icing method according to the ice thickness and conductor material. For areas with medium to thick ice, use resistance heating to quickly heat up and melt the ice layer; for thin ice areas or sensitive conductors, use electromagnetic induction heating to reduce local thermal stress.

[0083] 2. Power curve setting P(t) This defines the variation of heating power over time in each icing zone, used to control the heating rate and temperature rise curve. The power curve is represented as a function of time. , ,in For initial power, This is the power growth factor; by properly designing the power curve, de-icing can be completed within a specified time while avoiding overheating of the conductor.

[0084] in: Temperature rise curve: This curve represents the change in conductor temperature over time. It is the result of the combined effects of power input and the conductor's thermal properties (thermal conduction, heat dissipation, specific heat capacity, etc.).

[0085] Power curve: Represents the amount of heating power input to the conductor at different times, and is an externally applied energy control parameter.

[0086] 3. Work sequence arrangement: Sort the ice-covered areas according to thickness and area, and prioritize the treatment of areas with thicker ice or close to critical structures to ensure the efficiency and safety of ice melting operations.

[0087] 4. Safety threshold setting: Define the upper temperature limit during the ice melting process. T max The de-icing unit monitors parameters such as the maximum stress on the conductor and the insulation safety threshold in real time during operation to prevent damage to the conductor or equipment.

[0088] Example: In this scheme, it is assumed that a 10m long, 12mm thick icing area A appears on the aluminum core steel stranded wire, with an icing level of medium to thick; the strategy design unit selects resistance heating, and the power curve is as follows. P(t) = 500 + 50tW The operation time is set to 8 minutes, and the upper temperature limit is set to 80℃. If the ice thickness of adjacent area B is 5mm, and the ice thickness level is thin ice, then area A with the larger ice thickness will be melted first, followed by area B with thin ice. The ice melting strategy generated by the ice melting strategy design unit is then sent to the ice melting execution unit to guide the external device to perform targeted, efficient, and safe ice removal.

[0089] S23, the ice melting execution unit receives the spatial parameters, thickness and ice melting strategy of each ice-covered area generated by the ice melting strategy design unit, generates control commands for the external ice melting device, and moves the ice melting device to the target ice-covered area and implements the corresponding ice melting strategy; the ice melting execution unit drives the heating module to perform fixed-point ice melting on the target area according to the specified ice melting method (such as resistance heating or electromagnetic induction heating) according to the power curve P(t), and operates in the order of operation determined by the strategy.

[0090] During the de-icing process, the de-icing execution unit monitors the conductor temperature, stress state, and insulation safety parameters in real time using built-in sensors, comparing these parameters with the safety thresholds set by the strategy design unit. In case of any abnormality, the unit automatically adjusts the power or suspends the operation to ensure the safe operation of the conductor and equipment. Through control and monitoring, the de-icing execution unit precisely melts the ice in each section of the icing area, while simultaneously returning the operation data to the recording and feedback unit, achieving closed-loop operation management.

[0091] The abnormal handling of the ice melting execution unit of this invention is divided into automatic power adjustment or suspension of operation. For cases where the conductor temperature is slightly over the limit or the stress is slightly abnormal, the power is automatically adjusted; for cases where the conductor temperature is severely over the limit, the insulation condition is abnormal, or the sensor is faulty, the operation is suspended.

[0092] S24 utilizes a recording and feedback unit for centralized recording and management of data throughout the entire ice-melting operation process. This unit stores spatial parameters of each icing area, ice-melting strategy parameters (such as power curves, operation sequence, and operation time), and operation result information returned by the ice-melting execution unit, forming a complete operation data chain. The recording and feedback unit can also analyze historical operation data to provide a basis for subsequent strategy optimization, equipment maintenance, and operation planning.

[0093] A transmission line icing detection and de-icing device, comprising: Visible light image sensors are used to detect target transmission line areas and obtain optical image data of the transmission lines; Thermal imaging sensors are used to detect target transmission line areas and obtain thermal imaging image data of the transmission lines. The IS intelligent computing chip is used to detect icing and make decisions on de-icing based on data from visible light image sensors and thermal imaging sensors.

[0094] The IS intelligent computing chip includes: The icing identification and positioning unit is used to identify and locate areas with icing in the fused image of the power transmission line based on the data fused from the visible light image sensor and the thermal imaging sensor. The ice-melting strategy design unit is used to generate an ice-melting strategy based on the data from the ice-covering identification and positioning unit. The ice melting execution unit is used to receive the spatial parameters, thickness and ice melting strategy of each ice-covered area generated by the ice melting strategy design unit, generate control commands for the external ice melting device, and make the ice melting device move to the target ice-covered area and implement the corresponding ice melting strategy. The recording and feedback unit is used to centrally record and manage data from the entire ice melting operation process.

[0095] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for detecting and de-icing icing on transmission lines, characterized in that, The specific steps are as follows: S1, collects transmission line data and establishes an initial feature image through multi-source sensors; S2 is designed as an IS intelligent computing chip to realize icing detection and ice melting decision-making.

2. The method for detecting and de-icing icing on transmission lines according to claim 1, characterized in that, The specific steps of S1 are as follows: S11, the target transmission line area is detected by a visible light image sensor to obtain optical image data of the transmission line. D i Simultaneously, thermal imaging sensors are used to detect the same area, obtaining thermal imaging image data of the transmission line. D t ; S12, will D i and D t The data is input to the data acquisition unit for preprocessing. S13, finally according to the set weight parameters α and β By fusing the two types of data, a fused image of the transmission line at the site is obtained. D f .

3. The method for detecting and de-icing icing on transmission lines according to claim 1, characterized in that, The specific steps of S2 are as follows: S21, designed as an icing recognition and localization unit, is used to receive fused images. D f As input, D f After processing by the icing area perception model integrated within the unit, the field fusion of transmission lines is identified and located. D f There are areas covered with ice; S22, use the ice melting strategy design unit to generate the ice melting strategy; S23, the ice melting execution unit receives the spatial parameters, thickness and ice melting strategy of each ice-covered area generated by the ice melting strategy design unit, generates control instructions for the external ice melting device, and moves the ice melting device to the target ice-covered area and implements the corresponding ice melting strategy. S24 uses a recording and feedback unit to centrally record and manage data throughout the entire ice melting operation process.

4. The method for detecting and de-icing icing on transmission lines according to claim 3, characterized in that, The specific steps of S21 are as follows: S211, Design an RFDEM region feature depth extraction module, which realizes diversified extraction and fusion of input icing features through a multi-branch parallel processing mechanism; S212, Design an FMIM feature multi-scale interaction module, which realizes hierarchical extraction and fusion of input features of icy areas by using a multi-scale icing area feature interaction mechanism; S213, Design a MIFD multi-information fusion detector, which achieves accurate detection of the location information of the icing area through a four-level branch collaborative prediction mechanism; S214. Based on the RFDEM module, FMIM module, and MIFD detector, an icing area perception model IAPM is constructed to receive fused images of transmission lines. D f As input, it automatically identifies and locates. D f The ice-covered areas that exist in the middle.

5. A device for detecting and de-icing icing on power transmission lines, characterized in that, include: Visible light image sensors are used to detect target transmission line areas and obtain optical image data of the transmission lines; Thermal imaging sensors are used to detect target transmission line areas and obtain thermal imaging image data of the transmission lines. The IS intelligent computing chip is used to detect icing and make decisions on de-icing based on data from visible light image sensors and thermal imaging sensors.

6. The transmission line icing detection and de-icing device according to claim 5, characterized in that, The IS intelligent computing chip includes: The icing identification and positioning unit is used to identify and locate areas with icing in the fused image of the power transmission line based on the data fused from the visible light image sensor and the thermal imaging sensor. The ice-melting strategy design unit is used to generate an ice-melting strategy based on the data from the ice-covering identification and positioning unit. The ice melting execution unit is used to receive the spatial parameters, thickness and ice melting strategy of each ice-covered area generated by the ice melting strategy design unit, generate control commands for the external ice melting device, and make the ice melting device move to the target ice-covered area and implement the corresponding ice melting strategy. The recording and feedback unit is used to centrally record and manage data from the entire ice melting operation process.

Images